Justification Before Knowledge?

JustificationBefore Knowledge?

E.J. Coffman

University of Tennessee

[Word count: 8,908]

ABSTRACT.This paper assesses several prominent recent attacks on the view that epistemic justification is conceptually prior to knowledge. I argue that this view—call it the Received View (RV)—emerges from these attacks unscathed. I start with Timothy Williamson’s two strongest arguments for the claim that all evidence is knowledge (E→K), which impugns RV when combined with the claim that justification depends on evidence. One of Williamson’s arguments assumes a false epistemic closure principle; the other misses some alternative (to E→K) explanations of a putative fact about the evidence a particular subject has. Next, I neutralize each of Jonathan Sutton’s three recent arguments to the conclusion that any justified belief constitutes knowledge. Finally, I consider a recent analysis of justification due to Alexander Bird, according to whichjustified belief ispossible knowledge. I argue that Bird’s analysis delivers neither a sufficient nor (more importantly) a necessary condition for justification. [Word count: 149]

Introduction

The purpose of this paper is to assessseveral prominent recent attacks onthe view that epistemic justificationis conceptually prior toknowledge. I understand epistemic justification to be the kind of justification a belief has precisely when the belief would—if held in normal circumstances—earn “high marks” relative to theepistemicaim of having true beliefs.[1] At the risk of (perhaps gross) understatement, the view that the concept of knowledge is parasitic on that of (epistemic) justificationenjoys a large following among contemporary epistemologists. In his influential Knowledge and its Limits, Timothy Williamson observes that

Gettier’s counterexamples to [the traditional definition of knowledge as justified true belief] did not remove the idea that the concept of justification… would occur with the concept of belief in a more complex analysis of the concept of knowledge. Consequently, the concept of knowledge was assumed to be unavailable for use in an elucidation of the concept of justification…, on pain of circularity. (8-9)

In what follows, I’ll be arguing that this view about the relationship between justification and knowledge—call it the Received View(RV)—emerges unscathed from several prominent recent attacks. §1 tackles Timothy Williamson’s (2000) two strongest arguments forthe thesis that all evidence is knowledge (E→K), a claim that impugns RV when combined with the plausible view that justification depends (at least in part) on evidence. I show that, once clarified, the first of Williamson’s arguments for E→K assumes a false epistemic closure principle; the second argument depends on the false claim that E→K constitutes a uniquely good explanation of a putative fact about the evidence a particular subject has. In §2, I neutralize Jonathan Sutton’s (2005, 2007) three recent arguments for the conclusion that any justified belief is a case of knowledge. Finally, in §3 I consider a recent analysis of justification due to Alexander Bird (2007), according to which all and only cases of possible knowledge are justified beliefs. I argue that Bird’s analysis delivers neither a sufficient nor (more importantly) a necessary condition for justification.

1. Williamson’s Justification

In Knowledge and its Limits, Timothy Williamson makes two sustained attempts to establish E→K, a thesis he then combines with the view that any justified belief is based on evidence to yield the conclusion that justification is parasitic on knowledge.[2] Writes Williamson:

The present case for E = K [the thesis that all and only knowledge is evidence] is now complete. If evidence is what justifies belief, then knowledge is what justifies belief…

Could belief be epistemically justified except by evidence? Epistemic justification aims at truth in a sense—admittedly hard to define—in which pragmatic justification does not. It is far from obvious that any belief is justified in the truth-directed sense without being justified by evidence. It appears otherwise when evidence is conceived too narrowly, for then the evidence looks too scanty to justify all the beliefs which are in fact justified. But if anything we know can be evidence to anchor a chain of justification, as E = K implies, then evidence plausibly suffices for all truth-directed justification. An epistemically justified belief which falls shortof knowledge must be epistemically justified by something; whatever justifies it is evidence. An epistemically justified belief which does not fall short of knowledge is itself evidence, by E = K. If we are aiming at the truth, we should proportion our belief to the evidence.

E = K supports the plausible equation of truth-directed justification with justification by evidence, and therefore with justification by knowledge. On this view, if truth-directed justification is central to epistemology, so too is knowledge. (2000: 207-8)

Williamson’s argument to the conclusion that justification is parasitic on knowledge has two main premises: any justified belief is based on evidence (J→E), and all evidence is knowledge (E→K). For present purposes, I’ll simply grant J→E,[3] and will head off Williamson’s attack on RV by neutralizing his two strongest arguments for E→K. Before considering Williamson’s arguments for E→K, though, I want to bring out some of its initial implausibility, and distinguish it from another claim with which it might be confused.

Say that a known belief is a belief that qualifies as knowledge, and that a nonknown beliefis a belief that falls short of knowledge. E→K rules out a view that’s plausible in light of reflection on certain main aspects of the concept of evidence—viz., the view that the content of a nonknown belief could qualify as evidence the belief’s subject has. As many theorists note,[4]evidence is a functional concept, one having the form “item that plays the following roles: …”. Your evidence

▪ indicates to you that certain propositions are true;

▪ is “respected” by you when you’re thinking rationally;

▪ can be made available to others for use in adjudicating among competing views;

and(as we’ve noted)

▪ justifies your believing certain propositions.

This characterization of evidence seems to let at least some nonknown beliefs qualify as evidence. It seems that some justified false—or nonknown justified true (e.g., “Gettierized”)—beliefs could play all the roles just listed. Plausibly, some justified false—or nonknown justified true—beliefs could justify your believing certain propositions; indicate to you that certain propositions are true; be “respected” by you when you’re thinking rationally; and be made available to others (via testimony, e.g.) for use in adjudicating among competing views. Given the plausibility of the thought that a nonknown belief could play all the roles included in the above characterization of evidence, E→K should initially strike us as somewhat implausible.

I now want to distinguish E→K from another thesis with which it might be confused—viz.,

Evidential Externalism (EE): One’s total evidence doesn’t supervene on the contents of one’s nonfactive mental states.[5] That is, two subjects can differ with respect to their evidence without differing with respect totheir nonfactive mental states.

The important thing to see here is that E→K is stronger than EE: while E→K obviously entails EE, the latter doesn’t return the favor. To see that EE doesn’t entail E→K, consider a position that rejects E→K but says that a given mental state token provides its subject with evidence only if it is true/veridical. Such a position entails EE but not E→K. The upshot is that whatever plausibility EE enjoys does not automatically accrue to E→K. In addition to clarifying E→K, I think this point serves to underscore E→K’s initial implausibility.

We turn now to Williamson’s two strongest arguments for E→K. In the following passage, Williamson presents what I’ll call the Chain Reaction Argument:

If evidence required only justified true belief, or some other good cognitive status short of knowledge, then a critical mass of evidence could set off a kind of chain reaction. Our known evidence justifies belief in various true hypotheses; they would count as evidence too, so this larger evidence set would justify belief in still more true hypotheses, which would in turn count as further evidence… . The result would be very different from our present conception of evidence.(Williamson 2000: 201 [emphasis mine])

Before we can assess this argument, we’ve a preliminary hermeneutical chore: how should we interpret the italicized sentence in the passage?[6] Two readings present themselves: one involving doxastic justification—a feature had by beliefs, and expressed by locutions like ‘S has a justified belief that P’—, and one involving propositional justification—a feature had by propositions, and expressed by locutions like ‘P is justified for S’. Here’s the “doxastic justification” reading of the sentence:

Doxastic Justification Reading (DJR): One’s known evidence renders one possessed of variousjustified beliefs whose contents are true hypotheses.

And here’s the “propositional justification” reading:

Propositional Justification Reading (PJR): One’s known evidence renders various true hypotheses justified for one.

Which of DJR and PJR is the best interpretation of the italicized sentence?

Charity demands that we interpret the sentence as PJR, not DJR. DJR says that your simply having some known evidence suffices for your justifiedly believing various truths supported by that evidence. That’s implausible: it seems possible that you have some known evidence such that you don’t yetjustifiedly believe any of the truthssupported by that evidence. Perhaps you haven’t yet come to believe any of those truths. Or perhaps you believe some of those truths, but none of those beliefs is based on your evidence for it—and so, none of the beliefs is justified. In light of these possibilities, DJR is implausible. On the other hand, PJR says that your simply having some known evidence suffices for your having justification to believe various truths. Unlike DJR, PJR is quite plausible. Given PJR’s plausibility and DJR’s implausibility, charity demands that we interpret the italicized sentence as PJR, not DJR.

Hermeneutical chore done, we’re now ready to evaluate Williamson’s Chain Reaction Argument, which I reconstruct as follows:

Suppose, for reductio, that justified true belief suffices for evidence. Suppose also that you have some knowledge, and that it justifies various true propositions for you. By the reductio assumption, each of those truths now qualifies as evidence you possess. But some of those propositions don’t yet qualify as evidence you possess. So, the reductio assumption is false: justified true belief doesn’t suffice for evidence. Finally, for every other good cognitive status short of knowledge, there’s a parallel argument to the conclusion that the status in question does not suffice for evidence. Only knowledge qualifies as evidence.

What should we make of this argument?

Well, we can start by agreeing with Williamson that your knowledge can justify truths for you that don’t yet qualify as evidence you possess. A typical known belief of yours renders justified for you truths you’ve never entertained or considered.[7] But a proposition qualifies as evidence you possess only if you’ve at least entertained or considered it.[8] Thus, a true proposition’s being justified for you by your knowledge doesn’t suffice for its being evidence you possess. One of Williamson’s two premises is safe.

Unfortunately, we can’t say the same about the argument’s other premise:

Other Premise (OP): IF justified true belief suffices for evidence, THEN each true proposition your knowledge justifies for you also qualifies as evidence you possess.

To see that OP is false, consider

Consequence: IF (i) S knows that Pand (ii) S’s knowing P justifies Q (for S)and (iii) Q is true, THEN S has a justified true belief that Q.

According to Consequence, a true proposition’s being justified (for you) by something you know suffices for your having a justified true belief in that proposition. Now, notice that if Consequence is false, then OP is false. For suppose that justified true belief suffices for evidence, and that Consequence is false. Then the following three propositions are jointly compatible:

▪ S’s knowledge that Pjustifies the fact that Q(for S).

▪ S doesn’t yet justifiedly believe Q.

▪ Q is not yet evidence had by S.

Provided that Consequence is false, it’s possible that while justified true belief suffices for evidence, S’s knowledge justifies (for S) a true proposition that is not yet among S’s evidence. In short, if Consequence is false, it’s possible that OP have a true antecedent and a false consequent: if Consequence is false, OP is false. Finally, Consequence is clearly false. As we’ve noted, you might fail to entertain or consider some true proposition your knowledge justifies for you; a fortiori, you might fail to justifiedlybelieve some true proposition your knowledge justifies for you. Interestingly, the possibility that supports one of Williamson’s premises undermines the other. The Chain Reaction Argument fails.

The following passage expresses the second of Williamson’s two strongest arguments for E→K, what I’ll call the Filmed Drawing Argument:

Suppose that balls are drawn from a bag, with replacement… I watch [the draws] on film. For a suitable number n, the following situation can arise. I have seen draws 1 to n; each was red (produced a red ball). I have not yet seen draw n+1. I reason probabilistically, and form a justified belief that draw n+1 was red too. My belief is in fact true. But I do not know that draw n+1 was red. Consider two false hypotheses:

h: Draws 1 to n were red; draw n+1 was black.

h*: Draw 1 was black; draws 2 to n+1 were red.

It is natural to say that h is consistent with my evidence and that h* is not. In particular, it is consistent with my evidence that draw n+1 was black; it is not consistent with my evidence that draw 1 was black. Thus my evidence does not include the proposition that draw n+1 was red. Why not? After all, by hypothesis I have a justified true belief that it was red. The obvious answer is that I do not know that draw n+1 was red; the unsatisfied necessary condition for evidence is knowledge. (Williamson 2000: 200-01)

This argument strikes me as more promising than the first. Nevertheless, I think that it too fails.[9] For present purposes, I’ll simply grant Williamson’s Coherence Claim—viz., the (alleged)fact that [Draw n+1 was black] coheres with our subject’s evidence (at t).[10] Williamson contends that the best explanation of the Coherence Claim is

No Knowledge (NK): The subject doesn’t know that draw n+1 was red.

Williamson concludes that “the unsatisfied necessary condition for evidence is knowledge.” It’s at best unclear, though, whether NK really is the best explanation of the Coherence Claim. After all, there are multiple alternative explanations of the Coherence Claim that aren’t clearly inferior to NK.

Two technical terms will facilitate discussion of two alternative explanations I have in mind here. First, let warrant name the property that yields knowledge when conjoined with true belief. Second, letknowledge-level justification name the property a belief has iff the belief would constitute knowledge were it true and not “Gettierized”. It’s worth emphasizing that justification is logically weaker than knowledge-level justification: while the latter entails the former,[11] the former doesn’t entail the latter. To see that some justified beliefs aren’t also knowledge-level justified, consider (e.g.) your belief that you’ll lose the (large, fair) lottery you’ve entered. Plausibly, this belief would—if held in normal circumstances—earn “high marks” relative to the aim of having true beliefs. So, your belief that you’ll lose is justified.[12] But your beliefdoesn’t have knowledge-level justification: even if your belief were true and not Gettierized, you still might not know you’ll lose. Justification is weaker than knowledge-level justification.

Now, consider these two alternative (to NK) explanations of the Coherence Claim:

No Knowledge-Level Justification (NKJ): The subject doesn’t have a knowledge-level justifiedbelief that draw n+1 was red.

No Warrant (NW): The subject doesn’t have a warrantedbelief that draw n+1 was red.

Both NKJ and NW are true. As for NKJ, the subject’s belief that draw n+1 was red doesn’t constitute knowledge, despite its being both true and not Gettierized. As for NW, the subject’s belief doesn’t constitute knowledge, despite its being true. In light of these points, it’s far from obvious that neither NKJ nor NW explains the Coherence Claim as well as NK does. Thus, in the absence of considerations that show both NKJ and NW to be explanatorily inferior to NK here, Williamson can’t properly conclude that all evidence is knowledge from the Coherence Claim. Since such considerations haven’t been supplied, the Filmed Drawing Argument fails.

Before moving on, I should dispel a worry one might have about myobjection to Williamson’s Filmed Drawing Argument. My overall aim in this paper is to defend RV, the view that justification is conceptually prior to knowledge. Accordingly, I’ll want my objections to the anti-RV arguments I consider to be compatiblewith RV. But my objection to the Filmed Drawing Argument may seem toconflict with RV. I objected to the argument by noting two alternative (to E→K) explanations of the Coherence Claim: (i) the thesis that only knowledge-level justification suffices for evidence (E→JK-L), and (ii) the thesis that only warrant suffices for evidence (E→W). At a minimum, then,my objection provides raw material for an argument from the Coherence Claimto the disjunction of E→K, E→JK-L, and E→W. Crucially, though, I used the concept of knowledge in describing the concepts of knowledge-level justification and warrant. It looks, then, as though an argument to the indicated disjunction could be combined with J→E to yield the conclusion that justification is parasitic on knowledge after all. But if that’s right, then an argument to the specified disjunction is as lethal to RV as is an argument to E→K. Thus, my objection to the Filmed Drawing Argument seems to conflict with RV.